ResearchBenchmarksAdversarial Robustness Evaluation on MuJoCo Hopper LocomotionFollow3,513,182Target RewardPPO3,337,522.93,425,352.453,513,1823,601,011.55Jun 6, 2024Evaluation ResultsMethodMethodLinksTarget RewardClean RewardTargeted PGD RewardMax Reward (SA-RL)Reward (BIA-ILfD ours)Average Attack RewardPPODefense=No defenseDefense=No defense2024.063,513,182140,554167,354305,64231,9892,642TDRT-PPODefense=TDRTDefense=TDRT2024.063,513,1821,310,1911,324,3212,190,23414,8121,665